[BioC] Using RMA normalization on 3 microarray replicates - valid?

[James MacDonald]

Hi Kaitlin,

Incorporating the eighth timepoint into your experiment will be difficult, given that the chips used are likely quite different from the version 1.0 chips that you used for the other time points. Since you only used one timepoint with these chips you have aliased any biological differences with the chip type, so it will be impossible to determine if any differences are due to biology, technical differences or a combination of the two.

The fact that you have a large number of identical probeset values is an artifact of the median polish procedure that arises when you use three (or five) chips. This is not really the problem here, however. Although it is not ideal to estimate the expression values with such a small number of chips, the much larger problem is the chip-timepoint aliasing that I mentioned above.

Best,

Jim



James W. MacDonald, M.S.
Biostatistician
Douglas Lab
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826
>>> Kaitlin Louise Bergfield  04/01/10 4:47 PM >>>
Hello,

We are using Affymetrix Drosophila microarrays to investigate central
nervous system gene expression profiles at eight timepoints spanning
metamorphosis.  Each of the eight timepoints consists of three biological
replicate samples.  Unfortunately, our final eighth timepoint had to be
hybridized to version 2.0 arrays, while all our other samples were
hybridized to version 1.0 arrays.  We have found no way to normalize these
24 samples all together.  I have attempted to use RMA normalization on the 3
replicates from the final timepoint, but find when I do this that I end up
with vast numbers of identical values in the dataset.  These are highly
correlated (98-99%) with the raw values of the arrays, but I feel that with
only 3 replicates in this normalization I might be forcing artificial
conformity to a range of discrete normalized values.

My concern is that running RMA normalization on only 3 replicates is not a
valid use of the method.  Can anyone offer advice on the number of
replicates necessary to run RMA normalization, or if other methods are more
useful for this sort of analysis?

I have been using the following code:

cels <-dir("F:/Restifo Lab/Microarray files/A1 files",
pattern=".*.CEL", full.names=TRUE)
batch <- ReadAffy(filenames=cels)
eset <- rma(batch)
datamatrix <-exprs(eset)


Thank you very much,
Kaitlin Bergfield

-- 
Kaitlin Bergfield
Neuroscience Graduate Interdisciplinary Program
Brain Imaging, Behavior & Aging Lab
University of Arizona
kshupe at email.arizona.edu

    [[alternative HTML version deleted]]

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/bioconductor
Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues